透過您的圖書館登入
IP:13.59.218.147
  • 學位論文

多準則評分系統於合購網站主購之推薦

A Multi-criteria Rating System for Recommending Initiator on Group-Buying Websites

指導教授 : 胡宜中

摘要


隨著網路購物的蓬勃發展,「合購」已成為新興模式。此模式為一群具有相同購買需求的消費者,聚集在同一網路平台,透過網路群體的力量來與廠商議價,以達到降價或獲得相關優惠。資策會產業情報研究所(MIC)表示,全球及台灣的網路購物業者紛紛發展合購業務,我國合購市場規模於2010年達到71.6億元,2011年可望達到90億,由此可見合購於國內網購市場擁有很大的商機。隨著合購活動越趨盛行,合購網站快速地成長,使用者欲透過合購網站購買商品,一般透過搜尋功能找出合宜的主購。由於主購人數眾多,對使用者已著實造成資訊超載的現象。因此,如何協助使用者選擇最適合之主購係一亟待解決的重要問題。而推薦系統的主要目的即在於解決資訊超載的問題;然而,在電子商務領域中,過去研究多應用於推薦電影、音樂或書籍,但卻尚未應用於合購議題中,因此引發本研究之主要動機。 在過去,推薦系統之協同過濾模式多以單一準則評分為主,強調對待評項目給予整體評分;然而,使用者對主購的評價是屬於多準則決策問題,對主購的偏好可能反應在不同的準則上。因此除了單一準則評分系統外,使用能潛在提昇推薦正確率的多準則評分系統亦有其必要性。基於此,本研究即應用多準則評分推薦系統於合購網站主購之推薦。 經合購專家討論後,決定以「開團能力」、「信譽」、「回應程度」、「信任度」、「互動」等五項做為評估主購之準則。另考量準則間的相依性,故本研究使用網路分析程序法分析各個準則的相對權重,結果顯示「開團能力」與「信譽」為受訪者較為重視的準則,代表使用者希望主購具備能夠解決問題的知識或技術並且是可靠的,且會誠實地達到對團員的承諾。 本研究依此進一步提出考量準則權重之多準則評分系統,此與傳統多準則評分系統假設準則為等權重有很大的不同。研究結果顯示使用加權績效值的推薦系統在推薦正確率上較單一評分及傳統多準則評分系統為佳。同時,亦顯示本研究所提出之加權協同過濾式推薦系統在合購網站的主購推薦上的有效性。

並列摘要


As online shopping rapidly improves, "Group-buying" has become an emerging pattern. There are groups of consumers with the same needs, gathered in a network platform, negotiating with vendors to achieve lower prices through the power of network groups, or to obtain the relevant coupons. Institute for Information Industry (MIC) states that the online shopping industry in Taiwan reached 7.16 billion in 2010, and it’s expected to reach 9 billion by 2011, a sign of promising growth. With group-buying activities rising and websites expanding rapidly, users who want to purchase goods through group-buying websites need to search thoroughly to find appropriate ones. The number of the group-buy initiators has grown to the point of information overload. Therefore, how to help users choose the suitable initiator is a critical issue. The main purpose of the recommendation system is to solve this problem of information overload; However, recommender systems have been used for recommendation of films, news, books, and travels in the past, Group-buying online is a new form of e-commerce which recommender systems haven’t been applied to yet, and that leads to the principal motivation of this study. The vast majority of current recommender systems use a single criterion, such as a single numerical rating obtained from past purchases; however, the evaluation of the initiators is a multi-criteria decision. Multi-criteria rating can potentially increase recommendation accuracy. Based on this, this research applied multi-criteria rating system to the recommending initiator on group-buying websites. After much discussion, experts decided to use "capacity", "reputation", "response", "trust" and "interaction" as criteria for evaluation. To consider inter-dependencies, this research used a network analysis process to analysis the weight of each criterion, the results showing "capacity" and "reputation" as more important criteria for respondents, representing that most users hope the initiator can solve problems and is reliable, honest and will meet the commitment of the members. This research uses the weight value rating, which is different from the traditional multi-criteria rating that assumes weight values are the same. The results show that the precision of the multi-criteria recommender system is highest when we add the two methods above mentioned.

參考文獻


李依珊(2009),「影響消費者進行線上合購行為之因素研究」,中原大學資訊管理學系碩士論文。
吳晨帆(2005),「推薦策略與推薦資訊對線上推薦績效影響之研究」,中原大學資訊管理學系碩士論文。
陳亭光(2008),「基於使用者經驗之多準則評分遊戲推薦系統」,台灣大學資訊管理研究所碩士論文。
黃聆怡(2008),「電子口碑效果與網路合購行為之研究」,台灣大學管理學院國企所碩士論文。
閻瑞彥,周嘉俊(2008),「網站服務品質、商店形象、信任、知覺風險對顧客忠誠影響之研究-以電子商店為例」,電子商務研究,6(4),461-486.

延伸閱讀